A Framework for Anomaly Detection in Networks Using Machine Learning

Anomaly detection is the biggest challenge in real-world applications. It is important to detect such anomalies and take corrective measures to ensure the smooth functioning of networks. Anomaly detection removes functional threats and complications. It was noted that the research exhibited in this paper lacks strong pre-processing and feature extraction methods. The suggestion is to introduce suitable feature selection methods, the CICIDS2017 dataset, along with an anomaly detection system that includes ensemble learning of top-performing models. Feature selection methods can enhance the training quality of the data. A comparative evaluation is done here, and it was examined that the CICIDS2017 dataset gives more accuracy over the other datasets and is best suitable for anomaly detection in the networks. The proposal that this paper puts forth would all help the present deep learning and ML approaches to provide good anomaly detection.